We present three complete seasons and two half-seasons of SDSS r-band photometry of the gravitationally lensed quasar SBS 0909+532 from the U.S. Naval Observatory, as well as two seasons each of SDSS g-band and r-band monitoring from the Liverpool Robotic Telescope. Using Monte Carlo simulations to simultaneously measure the system's time delay and model the r-band microlensing variability, we confirm and significantly refine the precision of the system's time delay to \Delta t_{AB} = 50^{+2}_{-4} days, where the stated uncertainties represent the bounds of the formal 1\sigma\ confidence interval. There may be a conflict between the time delay measurement and a lens consisting of a single galaxy. While models based on the Hubble Space Telescope astrometry and a relatively compact stellar distribution can reproduce the observed delay, the models have somewhat less dark matter than we would typically expect. We also carry out a joint analysis of the microlensing variability in the r- and g-bands to constrain the size of the quasar's continuum source at these wavelengths, obtaining log[(r_{s,r}/cm) [cos{i}/0.5]^{1/2}] = 15.3 \pm 0.3 and log[(r_{s,g}/cm) [cos{i}/0.5]^{1/2}] = 14.8 \pm 0.9, respectively. Our current results do not formally constrain the temperature profile of the accretion disk but are consistent with the expectations of standard thin disk theory.